Centralized and Decentralized Methods for Multi-robot Safe Navigation
Xinbei Wang, Zexuan Yan, Licheng Zhong · 2022 International Conference on Machine Learning and Intelligent Systems Engineering (MLISE) · 2022
With the expansion of application scenarios to intelligent storage, unmanned vehicle distribution, etc., the multi-robot path planning (MRPP) problem has received widespread attention recently. Solutions to this problem are mainly classified into two categories: centralized methods and decentralized methods. In this article, two centralized methods (safe interval path planning and conflict-based search) and two decentralized methods (velocity obstacles method and nonlinear model-predictive control method) are simulated in their respective environments. The performance of each method is observed and evaluated. Preliminary comparisons are also made to discuss the application scenarios, advantages and disadvantages of each method by observing the outcomes of the four methods. In the simulation of centralized methods, conflict-based search performs better than safe interval path planning with a higher probability of success. When simulating decentralized methods, velocity obstacles method can work out a more efficient path for a robot than nonlinear model-predictive control. However, the robustness of velocity obstacles method can be poorer when applied to reality. The strengths and weaknesses should be balanced facing different demands.